| |
BGonline.org Forums
An old practice with a new twist? (long)
Posted By: Chuck Bower
Date: Thursday, 21 May 2015, at 12:10 a.m.
Way back when there weren't reasonable bots (let's say pre-1990) there was a practice among serious players to roll out, by hand, a position ~100 times. Part of the impetus was an attempt to determine the best choice. That was a nebulous goal given the statistical uncertainties of small samples sizes. A secondary (or was it primary?) value was to see how the game developed, which could affect one's conceptual understanding of the position and choice of plays.
Once bots came along, the compilation of results of hand rollouts (by comparison) became ridiculous. There was still value in the conceptual understanding / development side but the cost in time spent was still significant. The practice waned, if not nearly died out completely.
Today there is value in playing against a bot, recording the play-by-play (virtually free in the bot age) and analyzing the results. But there are issues (see below) with such types of games/matches. An alternative is to play against a human online and then have the bot analyze. OK, now let's look at the pluses and minuses.
1) "Realism" of the the opponent. Bots don't play like humans (and vice-versa). If your goal is to improve your human-human play, competing against a bot is less than ideal. a) bots make very few mistakes. b) bots assume their opponents are a smart as they are (and thus assume they make few mistakes). Here competing against online opponents wins hands-down.
2) Speed of play. Bots are very fast; humans typically much less so. You get in more decisions per unit time against bots. This is both a plus and a minus. The plus is obvious but the minus occurs when you let the bot determine the game's tempo. If you alter your speed to keep up with the bot (likely sub/unconsciously) then in most cases you damage your decision making.
3) Altruistic/golden-rule tendencies. You can't hurt a bot's feelings. If you leave the game without notice against a bot, no one feels pain (neither the droppee nor the conscientious dropper). That isn't the case with a human (online) opponent.
How does one's error rate compare when playing a bot and playing a human? That consideration has multiple knobs (see the tempo argument above) but one big one is that bots seldom lose their markets so simply taking all cubes, or better yet, taking all cubes where there wasn't a joker in the last sequence and only contemplating the joker situations, will lead to few cube errors. That typically doesn't happen with humans, because a) they don't know (as well) their market loser sequences, b) they don't know (as well) their actual equity, and in the case of top players, c) they consider the likelihood of opponent errors (this turn and next) into their cube decisions.
OK, enough of the preliminaries. My proposal is to use a bot as the platform but to play human-vs-human -- specifically you vs. yourself. Advantages:
1) you get more decisions per unit time compared to play vs. another human.
2) you don't get to take advantage of the bot's "don't lose the market against an equal opponent" cube tendencies.
3) you don't have the golden rule dilemmas (when playing against humans).
4) you are competing against the kind of opponent you're likely to encounter in a tournament, at least compared to playing against a bot.
What are the downsides?
1) you don't get the "thrill of victory", or at least it's tempered by an equal "agony of defeat". :) I don't see this as a downside if you are sufficiently disciplined.
2) you don't get the game theoretic components, such as bluff cubeturns. This may be overstated. Woolsey's Doubling Rule still applies: if you're not sure it's a take then double.
OK, my fingers are tired. Your thoughts?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.